Switch course ▾
Session 2. Policies That Teachers Can Use
Objective. Lead a group in writing five short, usable school policies on AI (acceptable use, assessment integrity, data protection, child safety, transparency), each on one page with clear rules, an owner, a supportive response to breaches and a review date, and test one of them with real teachers.
A policy is only as good as the Monday morning in which a tired teacher reads it. If she cannot tell in one minute what she may do, she will do what is easiest, or nothing, and the policy fails quietly. This lesson shows how to write rules with verbs, numbers and examples, how to build a ladder of supportive responses, and how to redesign assessed tasks so that integrity rests on good task design and evidence of understanding and not on guessing who used a tool.
Key ideas
Five short policies, not one long one
- One thick policy is signed and forgotten. Five one-page policies can be read in a staff meeting: acceptable use, assessment integrity, data protection, child safety, transparency.
- Each page has the same six parts: purpose, scope, rules (at most seven), who decides, what happens when a rule is broken, review date. Same shape makes them easy to learn and to audit.
- Your policies must sit inside national law and ministry regulations. This course does not quote specific articles; ask your education department or legal adviser which rules apply to personal data and child protection before you finish.
Acceptable use and assessment integrity
- Acceptable use answers: which tools are approved (the register from session 1), what may never be typed in (names, marks, photos, health or family details), and who checks every output before it reaches pupils or parents.
- Assessment integrity sorts tasks into three kinds: AI-free (exam, oral, in-class writing), AI-assisted with a declaration (pupil states what AI did), and AI-open (the task assumes AI). Say which kind each assessed task is, before pupils start.
- Check understanding through process: drafts, a three-minute oral explanation, a similar task in class. AI-text detectors can misjudge, especially writing by pupils who work in a second language, so a detector score must never be the only evidence.
Data protection and child safety
- Data protection in school language: collect only what you need, use it only for the stated purpose, say who can see it, say how long it is kept and how it is deleted. If a rule cannot be explained to a parent in two sentences, rewrite it.
- Child safety links to the school's existing child-protection procedure; do not create a second one. Add AI points: no unsupervised chatbot use by young pupils, filters are imperfect, and misuse such as fake images or voices of a classmate is treated as bullying and a safeguarding matter.
- Write who a pupil or teacher tells when something harmful appears, and promise that reporting is safe.
Transparency and a supportive response
- Transparency: publish a plain list of approved tools, what each is used for and what data it sees. Label materials that AI helped to draft. Pupils and parents should never learn about a tool by accident.
- When a rule is broken, the first response is a conversation and a fix, not a punishment. Reserve formal steps for repeated or serious breaches, and write which steps. Fear drives use underground.
- Test every policy: give it to five teachers, ask them to say in one minute what they may and may not do tomorrow. If they cannot, shorten it.
Three kinds of assessed task
AI-free
No tool allowed
- End-of-term test
- Oral explanation
- Writing in class
Declared
Tool allowed, pupil states how
- Take-home essay
- Reading log
- Project draft
AI-open
Task assumes a tool
- Compare and correct a draft
- Design a survey
- Fact-check a text
Rules people can follow
A usable rule has a verb, an object and a checker: 'Teachers enter no pupil names, marks or photos into any tool; the AI lead checks the register each term.' Test every rule with three questions. Could a new teacher follow it tomorrow? Could you see if it were broken? Does it say what to do instead? 'Use AI responsibly' fails all three. Add one example next to each rule, because examples travel better than principles.
Keep each page to seven rules or fewer and each rule under 20 words. Where the school already has a rule on phones or plagiarism, extend it rather than write a second one. Write the Uzbek and Russian versions together, not one after the other, so that the meaning is the same and neither language looks like a translation of an afterthought.
Task design and a ladder of responses
Integrity is easier to protect when the task asks for something a tool cannot supply alone: a local example from the pupil's own town, a comparison with a lesson discussed in class, drafts shown at two points, a three-minute oral explanation. Label each assessed task AI-free, declared or AI-open before pupils start, and let pupils see the label. Pupils working in a second language need special care: detector scores and a sudden change of style can unfairly point to them, so process evidence matters more.
Build a ladder of responses and write it down: first a private conversation and a chance to explain or redo; second a record and a call to parents; only third, for repeated or serious breaches, a formal step. A ladder tells pupils the school wants honesty and learning, and it protects teachers from improvising under stress.
Examples from the classroom
Twenty words
Scene. Jasur, a deputy head near Bukhara, collects the policies of four neighbouring schools and finds rules of 60 to 90 words that nobody remembers. With teachers he rewrites each rule to under 20 words, adds one example and a name of a checker. A test with six teachers follows: after a minute each can state three rules correctly, compared with one rule before. He sends the short version to every classroom.
Insight. Short rules with an example and a checker are remembered, followed and audited.
A task that shows the thinking
Scene. Malika teaches Grade 9 history in Namangan. She changes the essay 'Causes of the Aral Sea crisis' to a three-part task: a draft in class, a take-home version with a declaration, and a three-minute oral explanation to a partner. The pupils may use a tool for finding ideas, but they must explain every claim. She marks the explanation too, so pupils invest in understanding, and cheating becomes a poor strategy.
Insight. Redesigning the task makes evidence of understanding part of the assessment itself.
The second-language pupil
Scene. In a Russian-medium class in Tashkent, a pupil who changed from an Uzbek school two years ago submits a polished paragraph and a detector flags it. Dilnoza, the teacher, does not mention the score. She asks him to talk about the paragraph for two minutes, then to write a similar one in class. He succeeds. The school adds a line: a detector score is never evidence on its own.
Insight. Process evidence and a respectful conversation protect pupils from unfair suspicion.
Common pitfalls
- Writing rules in the language of lawyers, which teachers will not read.
- Banning every tool without saying how misuse would be noticed or why a ban is better than redesigning the task.
- Relying on detector scores or on a sudden change of style as proof of cheating.
- Punishing first, so that pupils hide their use rather than declare it.
Repair a Policy, Write an Integrity Rule
- Open the Activity Kit: the SAMPLE draft 'AI Policy v0' of Zarafshon School, a long text with eight defects (5 min).
- Pairs (10 min): find the defect in each of the eight lines and write what is wrong in a few words (vague, unenforceable, no owner, punishment first, and so on).
- Pairs (10 min): rewrite five of the lines as usable rules of at most 20 words, each with a named owner or checker.
- Pairs (10 min): classify the four assessed tasks in the kit as AI-free, AI-assisted with declaration or AI-open, and write the evidence of pupils' own understanding for each.
- Part B, pairs (15 min): write the one-page assessment integrity policy using the six-part template, including a declaration sentence pupils can sign and a supportive first response.
- Offline fallback: all work is on paper. To test a policy without devices, read it aloud to a neighbour pair, who must say back the rules in one minute.
Your facilitator will guide you through this activity in the session.
Questions to discuss
- Which defect was most common in your own school's documents, and why do such texts get written?
- For which assessed task would AI be useful if the task were redesigned rather than banned?
- What would a pupil who admits using AI hear from you, and does your wording invite honesty?
Reflect
- Which existing school rule on phones or plagiarism could be extended instead of writing something new?
- Who checks that a policy is followed, and how would you know it is not?
- Which of your pupils would be unfairly suspected by a detector?
Try it tomorrow
- Take one rule from your school and rewrite it in fewer than 20 words with an example and a checker.
- Label your next three assessed tasks AI-free, declared or AI-open, and tell pupils before they start.
- Write a three-step ladder of responses and read it to a colleague, asking if it feels fair.
Further reading and links
Five one-page policies with the same six parts, rules a teacher can follow tomorrow, assessment tasks labelled by AI role, and a first response that supports before it punishes.